Serine 232 and Methionine 272 Define the Ligand Binding Pocket in Retinoic Acid Receptor Subtypes*

The transcriptional response mediated by retinoic acid involves a complex series of events beginning with ligand recognition by a nuclear receptor. To dissect the amino acid contacts important for receptor-specific ligand recognition, a series of retinoic acid receptor (RAR) mutants were constructed. Transcriptional studies revealed that serine 232 (Ser 232 ) in RAR a and methionine 272 (Met 272 ) in RAR g are critical residues for the recognition of their respective receptor-selective analogs.Theidentification of these key amino acids in the ligand binding pocket is confirmed by the reported crystal structure of RAR g . Interestingly, the serine at position 232 in RAR a gives an explanation for the observed differences in the affinity of the naturally occurring ligand, all- trans -retinoic acid (t-RA), in this receptor compared with that for the other receptors, since hydrogen bonding would not be permitted between the hydroxyl of serine and the hydrophobic linker of t-RA. Using this model, a molecular mechanism for the transcriptional antagonism of a synthetic analog is suggested that involves an alteration in the structure of the receptor protein in the region around the AF2 domain in helix 12. Retinoic acid and its analogs (retinoids) regulate cellular proliferation and differentiation in higher The biological effects of these ligands are mediated by their binding to the retinoic acid receptors (RARs), 1 which are members of the superfamily of steroid-thyroid acetyltransferase gene product obtained from transfected cells using the chloramphenicol acetyltransferase enzyme-linked immunosorbent assays Prime 3 Prime, normalized for transfection efficiency by b -galactosidase activity. Results (EC are presented as the dose at which chloramphenicol acetyltransferase induction was half of the level observed for t-RA at 10 2 6

The transcriptional response mediated by retinoic acid involves a complex series of events beginning with ligand recognition by a nuclear receptor. To dissect the amino acid contacts important for receptor-specific ligand recognition, a series of retinoic acid receptor (RAR) mutants were constructed. Transcriptional studies revealed that serine 232 (Ser 232 ) in RAR␣ and methionine 272 (Met 272 ) in RAR␥ are critical residues for the recognition of their respective receptor-selective analogs.
The identification of these key amino acids in the ligand binding pocket is confirmed by the reported crystal structure of RAR␥. Interestingly, the serine at position 232 in RAR␣ gives an explanation for the observed differences in the affinity of the naturally occurring ligand, all-trans-retinoic acid (t-RA), in this receptor compared with that for the other receptors, since hydrogen bonding would not be permitted between the hydroxyl of serine and the hydrophobic linker of t-RA. Using this model, a molecular mechanism for the transcriptional antagonism of a synthetic analog is suggested that involves an alteration in the structure of the receptor protein in the region around the AF2 domain in helix 12.
Retinoic acid and its analogs (retinoids) regulate cellular proliferation and differentiation in higher eukaryotes. The biological effects of these ligands are mediated by their binding to the retinoic acid receptors (RARs), 1 which are members of the superfamily of steroid-thyroid hormone nuclear receptors. These receptors act as transcriptional enhancers that bind in a sequence-specific manner to their response elements (retinoic acid response elements) located within the promoter region of distinct retinoid-responsive genes. RARs activate transcription of those genes after ligand bound to the receptor induces conformational changes leading to activation (1)(2)(3)(4). A crystal structure of the ligand binding domain (LBD) of the closely related human retinoid-X receptor ␣ revealed a novel protein fold, an antiparallel ␣-helical sandwich, common to the members of this superfamily. Examination of this structure for a potential binding pocket for its ligand, 9-cis-retinoic acid, unveiled two large hydrophobic cavities within the N-terminal portion of the LBD in the vicinity of the s1,s2 ␤-hairpin and ␣-helix H5 (5). These findings are supported by results from ligand-photoaffinity binding experiments, which identified important residues for receptor-ligand binding in the LDBs of the glucocorticoid receptor (6) and RAR␣ (7,8). The amino acid sequence alignment of the nuclear receptor LBDs indicates that most of the residues identified by photoaffinity labeling or site-directed mutagenesis correspond to the 9-cis-retinoic acid receptor residues surrounding the putative 9-cis-retinoic acid binding pocket (9).
Utilizing site-directed mutagenesis and RAR-selective retinoids, we previously identified serine 232 (Ser 232 ) and threonine 239 (Thr 239 ) from the N-terminal portion of the LBD of RAR␣ and the corresponding alanine 225 (Ala 225 ) and isoleucine 232 (Ile 225 ) from RAR␤ to be essential for the recognition of retinoic acid and various analogs (11). In the present work, the amide linker region of Am-580 and the oxime linker region of BMS-185354 were used to precisely identify serine 232 (Ser 232 ) in RAR␣ and methionine 272 (Met 272 ) in RAR␥ as critical for the specific interaction of the receptor and retinoid. Experiments with an RAR␣ antagonist, BMS-185411, showed that Ser 232 is also involved in the transcriptional antagonist activity of this compound. In RAR␤, an alanine (Ala 225 ) at the position corresponding to Ser 232 in RAR␣ was shown to allow BMS-185411 to behave as an RAR␤-specific agonist. Analysis of several additional RAR␣or RAR␥-specific point mutants introduced along ␣-helix H3, H4/H5 and the ⍀ loop of RAR␤ detected a decrease in the transactivation activity of BMS-185411. The amino acids mutated in these experiments, according to the crystal model of RAR␥ LBD, do not directly interact with t-RA. This suggests * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. that the effect of these residues on the potency and selectivity of BMS-185411 could be due to intramolecular interactions within the receptor itself. These findings support the conclusion that the transcriptionally active ligand-receptor complex is a result of a series of direct and indirect interactions between the receptor and its selective ligand, which reflect the dynamic nature of both components.
The polymerase chain reaction protocol described by Cotecchia et al. (16) was used to construct several chimeric retinoic acid receptors. DNA fragments synthesized by polymerase chain reaction were cloned directly into pRAR␣0 using a combination of the BstEII, BsgI, and BclI cloning sites found in RAR␣. The corresponding cloning sites, which are not present in RAR␤ cDNA, were introduced into DNA amplified from this receptor using synthetic amplification primers. Amplification and recombinant primers were made using the Pharmacia Biotech Inc. Gene Assembler-4. To construct the chimera RAR␣(E)␤, the BsgI site at position 1194 in RAR␤ was removed using a similar protocol.
Retinoid Transactivation and Competition Analysis-Transfection of HeLa cells with DNA encoding wild type RAR␣, RAR␤, and the various chimeric receptors was performed as described (11,12). Retinoid efficacy was measured by the concentration of induced chloramphenicol acetyltransferase gene product obtained from transfected cells using the chloramphenicol acetyltransferase enzyme-linked immunosorbent assays kit (5 Prime 3 3 Prime, Inc., Boulder, CO). Activity was routinely normalized for transfection efficiency by ␤-galactosidase activity. Results (EC 50 ) are presented as the dose at which chloramphenicol acetyltransferase induction was half of the level observed for t-RA at 10 Ϫ6 M.
To evaluate antagonist activity of test retinoids, a transactivation competition assay was used. Transfected HeLa cells were treated 16 h with 10 Ϫ7 M t-RA, with or without BMS-185411 at concentrations of 10 Ϫ8 to 10 Ϫ5 M. These cells were then harvested and crude extracts prepared. Chloramphenicol acetyltransferase protein level was assayed as described above. The antagonist activity of BMS-185411 is measured as an IC 50 for the inhibition of transactivation produced by t-RA.
Binding of synthetic ligands to the receptors was accomplished by a competition binding assay described in Ref. 11.
DNA and Protein Analysis-DNA sequence analysis and protein structure predictions were performed using GeneWorks 2.3 DNA-protein analysis software from IntelliGenetics, Inc.

Characterization of Retinoic Acid
Receptor ␣-, ␤-, and ␥-Specific Activities Using Receptor-selective Retinoids-Wild type RAR␣, -␤, and -␥ transiently transfected into HeLa cells were used to establish the profile of receptor-specific transactivation responses utilizing t-RA and the receptor-selective synthetic retinoids Am-580, BMS-185411, and BMS-185354 (Fig. 1). In these experiments, Am-580 was found to be an RAR␣-selective agonist, displaying an EC 50 of 3.2 nM for this receptor. BMS-  185411 showed specific agonist activity for RAR␤ with an EC 50 of 34 nM, and BMS-185354 selectively activated RAR␥ with an EC 50 of 28 nM ( Table I).
Effects of Serine 232 in RAR␣ on Specific Interactions with Receptor-selective Ligands-Utilizing the chimeric RAR, RAR␣(nE)␤, where the N-terminal portion of RAR␤ domain E was subcloned into an RAR␣ background by polymerase chain reaction-assisted site-directed mutagenesis, two sets of residues, serine 232 (Ser 232 )/alanine 225 (Ala 225 ) and threonine 239 (Thr 239 )/isoleucine 232 (Ile 232 ) in RAR␣ and RAR␤ were found to be essential for receptor-ligand-specific interactions (11).
Further investigation of the role of each residue alone was accomplished by a second series of mutants, RAR␣(nE-S225)␤, RAR␣(nE-T232)␤, RAR␣(nE-S232)␤, and RAR␤(S225). In the RAR␣(nE-S225)␤ mutant, alanine 225 in RAR␣(nE)␤ was substituted by a serine as found in RAR␣ (Fig. 2). In transactivation experiments, the RAR␣-selective agonist, Am-580, activated RAR␣(nE-S225)␤ with a profile similar to wild type RAR␣, with an EC 50 of 2.5 nM. Both the RAR␤-selective agonist, BMS-185411, and the RAR␥-selective agonist, BMS-185354, showed little or no transactivation activity with this mutant (Table I). These results suggested that Ser 232 in RAR␣ is the residue that is primarily responsible for this receptor's characteristic activities with selective retinoids like Am-580, which contains an amide linker. This conclusion was further supported by results of transactivation experiments with Am-580 and RAR␣(nE-T232)␤, where isoleucine 232 (Ile 232 ) in RAR␣(nE)␤ was replaced by the threonine found in RAR␣ (Fig.  2). This substitution did not alter the transactivation profile observed for the double mutant, RAR␣(nE-S225/T232)␤, or the single mutant, RAR␣(nE-S225)␤, indicating that only the ser-ine residue at position 225 is responsible for interaction with the amide linker of Am-580 (Table I). In addition, substitution of isoleucine 232 (Ile 232 ) by serine instead of threonine (mutant RAR␣(nE-S232)␤) showed an activity profile identical to that of the wild type RAR␤, indicating that addition of a second serine at this position will not alter the transactivation selectivity for Am-580. Finally, analysis of a single point mutant of RAR␤, RAR␤(S225), showed that the single amino acid substitution of alanine 225 (Ala 225 ) by serine resulted in a mutant receptor that responded to both t-RA and Am-580 with an identical profile to wild type RAR␣ and showed only weak transactivation with the RAR␥-selective, BMS-185354 (Table I). The EC 50 of 1.8 nM for Am-580 with RAR␤(S225) verified that Ser 232 in RAR␣ is the only amino acid which selectively interacts with the amide linker of the compound.
Methionine 272 Is Responsible for RAR␥-specific Transactivation Activity in Response to RAR␥-selective Retinoids-Examination of amino acid differences among RAR␣, -␤, and -␥ within the first 100 residues of the N-terminal portion of domain E revealed only a limited number of amino acid differences (Fig. 2). These suggested that the receptor chimera, RAR␣(nE)␤, could be used as a host to create five constructs targeted to identify the amino acids involved in specific RAR␥ligand interactions.
In a series of transactivation experiments, neither the RAR␤ arginine 212 substitution with RAR␥ glutamine (Gln 212 ) (RAR␣(nE-Q212)␤) nor the RAR␤ isoleucine 246 substitution with RAR␥ serine (Ser 246 ) alone (RAR␣(nE-S246)␤) or in combination (RAR␣(nE-Q212/S246)␤) was sufficient to convert a ligand-selective response from the RAR␤ type into the RAR␥ type using the RAR␥-selective BMS-185354. This suggests that both Gln 212 and Ser 246 are not contact amino acids between FIG. 2. Amino acid alignment of RAR␣, -␤, and -␥ LBD. Amino acids indicated by black arrows represent residues that have been mutagenized in this report and showed a direct effect on receptor-ligand interactions. Gray arrows indicate amino acids that are proposed to participate in intramolecular interactions under ligand binding conditions. Connected bars under the amino acid alignment correspond to ␣-helical regions identified in RAR␥ by x-ray crystallography (10). Numbers above the arrows refer to the positions of the amino acid residues in the RAR␣ (black), RAR␤ (green), and RAR␥ (red) sequence for comparison. RAR␥ and its selective ligand (Table I). The replacement of the RAR␤ isoleucine 263 (Ile 263 ) by methionine, which corresponds to the RAR␥ methionine 272 (Met 272 ) (Fig. 2), resulted in chimera RAR␣(nE-Q212/S246/M263)␤. In transactivation experiments, all three selective retinoids showed an EC 50 for this mutant receptor similar to that for the wild type RAR␥ (Table  I). This suggests that the methionine may be responsible for the RAR␥-specific interactions observed in this class of receptor-selective ligand. The results of experiments with RAR␣(nE-M263)␤, in which a single Ile 263 to methionine substitution effectively converted the chimeric receptor to an RAR␥ isotype, confirms this conclusion. Both Am-580 and BMS-185354 showed an EC 50 for the single mutant similar to that for wild type RAR␥ (Table I). Taken together, these results strongly suggest that a single isoleucine-methionine substitution is sufficient to convert the chimeric receptor, RAR␣(nE)␤, which shows an RAR␤-specific type of transactivation, into a receptor with an RAR␥-specific type of transactivation.
RAR-selective Transcriptional Antagonists Interact with Serine 232 in the Ligand Binding Pocket-Analysis of the results of transactivation experiments revealed that BMS-185411 acts as an RAR␤-specific agonist with an EC 50 of 34 and 54 nM for the wild type and chimeric RAR␣(nE)␤ receptors, respectively (Table I). In competition assays, however, this retinoid behaved as a specific antagonist for RAR␣, displaying an IC 50 of 400 nM (Table II). Furthermore, BMS-185411 is an antagonist for the mutants RAR␣/␤(nE)␣ and RAR␣(nE-S225/T232)␤, as well as for the chimeric receptors RAR␣(nE-S225)␤ and RAR␤(S225) with IC 50 values of 233, 220, 217, and 433 nM, respectively (Table II). The antagonist activity seen with RAR␣(nE-S225)␤ and RAR␤(S225) suggested that the Ser 232 residue in RAR␣ must also be involved in antagonist activity and selectivity. This conclusion is further supported by the results of competition experiments using RAR␣(nE-T232)␤ in which BMS-185411 did not act as an antagonist (Table II). Based on these results, we conclude that Ser 232 in RAR␣ and Ala 225 in RAR␤ are both selective contact amino acids which are responsible for both agonist and antagonist selectivity for the synthetic retinoids used in this report.

TABLE II
Results of retinoic acid competition assay Comparison of transcriptional antagonism of BMS-185411 for wild type RAR␣, RAR␤, RAR␥, and chimeric receptors. Transactivation was stimulated by 100 nM t-RA and then competed with increasing concentrations of BMS-185411. The level of competition is expressed as IC 50 , which is the concentration of each compound which inhibits 50% of the transactivation induced by t-RA. Numbers represent the mean of a minimum of three independent experiments.

DISCUSSION
The retinoic acid receptors are ligand-dependent transcription factors that regulate the expression of genes involved in cell growth, differentiation, and development. Yet, despite the large number of biological activities mediated by the receptors, relatively little is known about the precise interaction of the receptor protein and its naturally occurring ligand, t-RA. This study was undertaken to clarify the role of some of the amino acid contacts that were described in a previous report (11) and identified in the x-ray crystal structure (10).
In our previous report, two amino acid residues from the N-terminal region of the LBD, Ser 232 and Thr 239 in RAR␣, and the corresponding Ala 225 and Ile 232 in RAR␤, were shown to be critical for receptor-specific interactions with t-RA and various receptor-selective retinoids (11). Here, further analysis of these key amino acids using site-directed mutagenesis revealed that Ser 232 in RAR␣ and Met 272 in RAR␥ (Fig. 2) are critical for receptor-selective ligand interactions. The serine residue interacts with the amide linker portion of Am-580 in RAR␣, and the methionine residue interacts with the oxime linker in an RAR␣-selective retinoid (Fig. 3). Defined in this way, these receptor-selective contact amino acids are located in the ligand binding pocket and must modulate the binding of naturally occurring retinoids as well.
Using the results generated here for synthetic ligands, Ser 232 in RAR␣ may participate in a hydrogen bond with the amide linker of Am-580 and its analogs, which links the compound to ␣-helix H3 (Fig. 3). In RAR␤ and RAR␥, alanine residues at positions 225 and 234, respectively, correspond to Ser 232 in RAR␣. The presence of a hydrophilic linker, as seen in Am-580, in close proximity to a hydrophobic amino acid (Ala 225 or Ala 234 ) may result in negative interactions between the receptor and retinoid manifested as decreased affinity of the compound for both RAR␤ and ␥ (Fig. 3 and Table III). In contrast, mutant receptors containing Thr 239 alone were found to have no effect on the activity of the RAR␣-selective agonist, Am-580, and its analogs (Table I). This finding is supported by the results of experiments with the chimeric receptor RAR␣(nE-S232)␤, where the Thr 239 was substituted by serine.
The strategy outlined above was also used for the identification of RAR␥-selective amino acid contacts with receptor-selective ligands. Several mutant receptors were made, particularly an isoleucine 263 substitution for methionine in RAR␣(nE)␤, which produced RAR␣(nE-M263)␤. This mutation was found to be completely sufficient to convert the RAR␤ type of response of RAR␣(nE)␤ with BMS-185354 into the RAR␥ type. Therefore, it is likely that the methionine residue at position 272 is exclusively responsible for the RAR␥ selectivity of BMS-185354 (Table I). Methionine is generally considered to be a hydrophobic amino acid. However, the presence of a sulfur atom in the side chain of this amino acid could account for weak hydrogen bonding (18) between this amino acid and the hydroxyl substituent of BMS-185354 ( Fig. 1 and compound 2 in Ref. 11). On the other hand, the overall strong hydrophobic properties of methionine can also explain why there is only a moderate negative effect of RAR␥-selective retinoids on RAR␤, which contains an isoleucine residue at a position corresponding to Met 272 (Figs. 2 and 3). These observations are supported by conclusions from the crystal structure of RAR␥, where interactions between Met 272 and t-RA were proposed (10). The Met 272 is located on ␣-helix H5 (Fig. 2), which, according to the model from x-ray analysis, lies above the ligand (10). This arrangement also explains the results of experiments where the racemate (R,S)-1 of 6- (5,5,8,8-tetramethyl-5,6,7,8-tetrahydro-2-naphthyl) hydroxymethyl-2-naphthalene carboxylic acid and both its purified S-1 and R-1 enantiomers were used (19). All three compounds showed RAR␥ selectivity with the S-1 enantiomer being 10-fold more potent than R-1. In this enantiomer, the hydroxyl group is more favorably positioned with respect to Met 272 and, thus, Met 272 appears to be the only amino acid required for receptor-specific recognition.
Recently, another group of synthetic retinoids were discovered that act to reverse the effect of retinoic acid in transactivation assays. BMS-185411 is an example of such a transcriptional antagonist, which is shown here to interact with Ser 232 (Table II). We propose that the hydrogen bond between the hydroxyl group of Ser 232 and the amide linker of BMS-185411 positions this compound adjacent to ␣-helix H3 and may also force the phenyl group of this compound into close proximity to ␣-helix H12, which is important for transactivation activity (Fig. 3) (21, 22). The crystal structure of RAR␥ suggests that when t-RA binds to the receptor, ␣-helix-H12 closes the entry opening by swinging upward and interacting with ␣-helix-H4 (10,17,20). The ␣-helix-H12 is then stabilized in this position by a salt bridge between glutamic acid 412 (Glu 412 ) from H12 and lysine 264 (Lys 264 ) from ␣-helix-H4 (10). Modeling studies of the receptor with this ligand suggest that the phenyl ring of BMS-185411 may interfere with the proper alignment of ␣-helix-H12 leading to the inactivation of this receptor, possibly by interfering with the function of the AF-2 transactivation domain. In RAR␤, the lack of the hydrogen bond between alanine 225 and the amide linker of BMS-185411 may permit flexibility of the compound in the ligand binding pocket of this receptor thus avoiding interference with ␣-helix-H12 and leading to selective agonist activity ( Fig. 3 and Table I).
A similar explanation could also be applied to RAR␥, which contains an alanine residue in a position corresponding to RAR␤-Ala 225 and a methionine at position 272. Using the  RAR␣(nE-M263)␤ mutant, BMS-185411 was approximately 10 -15-fold less potent than the same compound with RAR␣(nE)␤ or wild type RAR␤ (Table I). This suggests that when BMS-185411 moves within the ligand binding pocket to avoid conflict with the ␣-helix-H12, the dimethyl groups from the tetramethyltetrahydoxynaphthalene portion of the compound interact with ␣-helix-H5 at position Met 272 (Fig. 3). The corresponding Ile 263 in RAR␤ is smaller and more hydrophobic than methionine; therefore, it can better accommodate hydrophobic dimethyl groups in its vicinity allowing for more potent agonist activity with RAR␤ (Table I).
The results obtained here for synthetic ligands can be used to explain the pattern of binding observed for the naturally occurring ligand of these receptors, t-RA. When bound to RAR␤, t-RA encounters an entirely hydrophobic environment (Fig. 4), which is reflected in the low apparent K d with Ile 263 and Ala 225 at the two key positions (Table III). In RAR␥, t-RA encounters Met 272 , a partial hydrophobic residue, and Ala 234 at these two positions giving an intermediate apparent K d . Finally, in RAR␣, an Ile 270 and Ser 232 generates hydrophilic character in the binding pocket and increases the apparent K d accordingly.
In summary, these results suggest that ligands interact with the amino acids in the ligand binding pocket in a manner that involves subtle positioning between the top of the pocket, represented by the methionine at position 272, and the bottom of the pocket. Our studies with both receptor-selective agonists and antagonists suggest that the interaction between RARs and their selective retinoids has a dynamic nature where both protein as well as ligand affect the final conformation to form a transcriptionally active complex. The functional analysis detailed here provides experimental confirmation of the RAR␥ LBD x-ray crystal structure. In addition, the identification of specific amino acid contacts within the ligand binding pocket can be exploited for the design of compounds of pharmacologic importance aimed at increasing the biological response and eliminating unwanted side effects.